Skip to content
Get packs

Agent Observability

Add tracing, token, latency and cost monitoring to LLM agents for debugging and reliability.

Source Download skill
Repository
sickn33/agentic-awesome-skills
Stars
49,023
Licence
MIT
Files
2 files

What it does

This skill teaches an AI agent how to help you instrument an LLM-powered application so you can see what it is doing: how long each step takes, how many tokens it uses, what it costs, and where errors happen. It provides Prometheus metric definitions, OpenTelemetry tracing patterns, and guidance on structured logging, SLOs and redacting personal data before telemetry leaves your system. The files are documentation and code examples; the agent adapts them to your codebase.

Great for

  • Engineers running AI agents or LLM features in production who need to debug slow or failing requests
  • Teams that want visibility into token usage and cost per agent, model or tool
  • Developers adding tracing to an agent framework using OpenTelemetry and Prometheus

Skip it if

  • Non-developers without access to the application code or its deployment, since the skill assumes you can wire up metrics and exporters
  • Environments where prompts or user data cannot be sent to a third-party telemetry backend, because the guidance itself warns that telemetry export is a data disclosure and prompts, credentials and user IDs should be omitted or approved first
  • Use against systems you are not authorised to instrument, as it involves connecting to production collectors and environments

Try asking

  • “Add Prometheus latency and token-usage metrics to my agent's LLM and tool-call code”
  • “Wrap my OpenAI chat completion calls with OpenTelemetry spans that record model, tokens and duration”
  • “Define a cost counter per model and agent and explain how to alert on spend”
  • “Help me set up PII redaction so prompts and user IDs are not exported to my tracing backend”
Step-by-step guide: how to use Agent ObservabilityInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 2 original files, the MIT licence and an attribution note.

    agent-observability.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip agent-observability.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip agent-observability.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Tell the agent which provider and model you use so it adapts the OpenAI example to your SDK
  • Ask for a review of what gets exported before enabling tracing in production, including prompt and user data handling
  • Use the bounded label values the skill recommends to avoid high-cardinality metrics that become expensive

What's inside

  • references/details.md
  • SKILL.md

Redistributed unmodified from sickn33/agentic-awesome-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.